摘要
根据BP(Back-Propagation)神经网络原理,以气、液表观流率为输入变量,液相传质系数为输出变量,建立神经网络模型,并利用改进的LM(Levenberg-Marquardt)算法对网络进行了训练和优化。结果表明,BP神经网络能够较好地预测滴流状态下H2O吸收CO2液相传质系数。
Based on the principle of BP（Back-Propagation）neural network, a neural network model was established, with gas and liquid velocities as inputs, liquid-side mass transfer coefficient as output. By using Levenberg-Marquardt（LM）algorithm, the net was trained and optimized. Results showed that BP neural network could predict liquid-side mass transfer coefficient for CO2 absorbed by H2O in the trickled condition.
出处
《长春工业大学学报》
CAS
2006年第4期283-285,共3页
Journal of Changchun University of Technology
基金
吉林省科技厅科学基金资助项目(19980564)
关键词
液相传质系数
BP神经网络
liquid-side mass transfer coefficient
BP neural network.
